CORNELL UNIVERSITY MATHEMATICS DEPARTMENT SENIOR THESIS Continued Fractions A THESIS PRESENTED IN PARTIAL FULFILLMENT OF CRITERIA FOR HONORS IN MATHEMATICS Yu Tung Cheng May 2007 BACHELOR OF ARTS, CORNELL UNIVERSITY THESIS ADVISOR(S) Peter Kahn Department of Mathematics Abstract This paper examines some properties and theorems of continued fractions. The definitions, notations, and basic results are shown in the beginning. Then peri- odic continued fractions and best approximation are discussed in depth. Finally, a number of applications to mathematics, astronomy and music are examined. Keywords: Continued fractions AMS Subject Classification: 11A55 2 Acknowledgements I would like to thank my thesis supervisor Professor Peter J. Kahn, without whose direction, teaching and counsel this paper would not exist. I would also like to thank my friends who gave me comments on this paper. Last but not least, I would like to thank my parents back home for their support. 3 Contents 1 Definitions 7 2 The Representation of Rational Numbers by Continued Frac- tions 10 3 Classical Theorems 14 4 Periodic Continued Fractions 19 5 Best Approximation 26 6 Applications 32 6.1 Quadratic Equations . 32 6.2 Calendar Construction . 34 6.3 Astronomy . 36 6.4 Music . 45 Appendix A Codes 47 Appendix B Some Miscellaneous Continued Fractions 49 Appendix C References 53 4 List of Tables 1 Solving x2 − 6x + 8 = 0, using 3 as initial value . 33 2 Solving x2 − 6x + 8 = 0, using 17 as initial value . 33 3 Solving 34x2 + 3254x − 549 = 0, using 9452.183 as initial value . 33 4 The first few convergents of the error between tropical and com- mon year . 34 5 Gear ratios of the solar system used by Huygens . 36 6 Gear ratios of Uranus and Neptune . 36 7 Gear ratios of dwarf planets . 40 8 Gear ratios of the planetary system of Gliese 876 . 41 9 Gear ratios of the planetary system of Mu Arae . 42 10 Gear ratios of the Saturnian system . 43 11 Gear ratios of the Uranian system . 44 3 12 The first few convergents of log( 2 )/log(2) . 46 5 List of Figures 1 Photograph of Huygen’s automatic planetarium [27] . 37 2 Sketch of Huygen’s automatic planetarium (front) [27] . 38 3 Sketch of Huygen’s automatic planetarium (back) [27] . 39 4 Eris, Pluto, Ceres, Moon and Earth [53] . 40 5 Gliese 876 planetary system [21] . 41 6 Mu Arae [35] . 42 7 The Saturnian system (photographic montage) [46] . 43 8 Octave [36] . 45 9 Perfect fifth [38] . 45 6 1 Definitions A continued fraction is an expression in the form b1 a0 + (1) b2 a1 + b3 a2 + b4 a3 + b5 a + 4 ... In general, the numbers a1, a2, a3, ... , b1, b2, b3, ... may be any real or complex numbers or functions of such variables. The number of terms can be finite or infinite. The discussion here will be restricted to simple continued fractions or regular continued fractions, which have the form 1 a0 + (2) 1 a1 + 1 a2 + 1 a3 + 1 a + 4 ... where a0 is an integer and the terms a1, a2, ... are positive integers. A finite simple continued fraction has only a finite number of terms, with the form 1 a0 + (3) 1 a1 + 1 a2 + 1 ... + 1 an−1 + an More precisely, this is an n-th order continued fraction. An n-th order con- tinued fraction has n + 1 elements. The terms a0, a1, a2, a3, ..., an are called the partial quotients of the continued fraction. 7 For simplicity, the simple continued fraction (2) can be called as infinite con- tinued fraction while the finite simple continued fraction (3) can be called as finite continued fraction. For technical convenience, an infinite continued fraction (2) can be represented as [a0, a1, a2, ...], while a n-th order continued fraction (3) can be represented in the form [a0, a1, a2, ..., an]. A segment of the finite continued fraction [a0, a1, a2, ..., an] is sk = [a0, a1, a2, ..., ak] (4) where 0 ≤ k ≤ n. Similarly, sk can be a segment of the infinite continued frac- tion for arbitary k ≥ 0. A remainder of the finite continued fraction [a0, a1, a2, ..., an] is rk = [ak, ak+1, ..., an] (5) where 0 ≤ k ≤ n. Similarly, rk = [ak, ak+1, ak+2, ...] (6) can be a remainder of the infinite continued fraction for arbitary k ≥ 0. Every finite continued fraction [a0, a1, a2, ..., an] is the result of a finite num- ber of rational operations on its elements, and can be represented as the ratio of two polynomials P (a0, a1, a2, ..., an) Q(a0, a1, a2, ..., an) in a0, a1, a2, ..., an with integral coefficients. As long as a0, a1, a2, ..., an are nu- merical values, the given continued fraction can be represented as a normal p fraction q . This representation is not unique, a way to solve this is as follows by induction. Notice that 1 [a0, a1, a2, ..., an] = [a0, r1] = a0 + (7) r1 8 Represent r1 as p0 r = , (8) 1 q0 then q0 a p0 + q0 [a , a , a , ..., a ] = a + = 0 (9) 0 1 2 n 0 p0 p0 r1 is a finite segment that has already been calculated by induction. So it is possible to set p [a , a , a , ..., a ] = (10) 0 1 2 n q p0 r = [a , a , ..., a ] = , 1 1 2 n q0 As r1 is also a continued fraction itself, this process can continue by the equations 0 0 p = a0p + q (11) q = p0 Denote by pk/qk the canoncial representation of the segment sk = [a0, a1, a2, ..., ak] and it is the k-th order convergent of the continued fraction. For an kth-order continued fraction γ, obviously pk/qk = γ. 9 2 The Representation of Rational Numbers by Continued Fractions p A rational number is a fraction of the form q where p and q are integers with q 6= 0. Here is an example: 93 23 1 1 1 1 = 2 + = 2 + = 2 + = 2 + = 2 + 35 35 35 12 1 1 1 + 1 + 1 + 23 23 23 11 1 + 12 12 1 1 = 2 + = 2 + = [2, 1, 1, 1, 11] (12) 1 1 1 + 1 + 1 1 1 + 1 + 12 1 1 + 11 11 It turns out that any rational number can be represented by a continued fraction, as stated by the following theorem. Theorem 2.1 Any finite simple continued fraction represents a rational num- ber. Conversely, any rational number p/q can be represented as a finite simple continued fraction. Proof The proof will be shown in Theorem 5.2. Theorem 2.2 Any rational number p/q can be represented as a finite simple continued fraction in which the last term can be modified so as to make the number of terms in the expansion either even or odd. Proof Suppose that p = [a , a , ..., a ] (13) q 0 1 n If an = 1, then 1 1 = (14) 1 an−1 + 1 an−1 + an 10 and it is possible to represent p = [a , a , ..., a , a + 1] (15) q 0 1 n−2 n−1 If an > 1, then 1 1 = (16) an 1 (a − 1) + n 1 and it is possible to represent p = [a , a , ..., a , a − 1, 1] (17) q 0 1 n−2 n−1 QED To avoid this ambiguity, the last partial quotient an of any finite continued fraction will be restricted to be greater than 1 from now on. Now, consider the following: 35 35 1 1 1 1 = 0 + = 0 + = 0 + = 0 + = 0 + 93 93 93 23 1 1 2 + 2 + 2 + 35 35 35 12 1 + 23 23 1 1 1 = 0 + = 0 + = 0 + 1 1 1 2 + 2 + 2 + 1 1 1 1 + 1 + 1 + 23 11 1 1 + 1 + 12 12 12 11 1 = 0 + = [0, 2, 1, 1, 1, 11] (18) 1 2 + 1 1 + 1 1 + 1 1 + 11 11 93 Compare this with 35 = [2, 1, 1, 1, 11] in (12). In fact, it can be generalized to the following theorem: p Theorem 2.3 For integers p, q where p > q, q = [a0, a1, a2, .., an] if and only q if p = [0, a0, a1, a2, ..., an]. Proof p If p > q > 0, then > 1 and q p 1 = [a0, a1, a2, ..., an] = a0 + (19) q 1 a1 + 1 a2 + 1 a3 + 1 a4 + 1 ... + an p where a is an integer> 0. The reciprocal of is 0 q q 1 1 = = (20) p p 1 a0 + q 1 a1 + 1 a2 + 1 a3 + 1 a4 + 1 ... + an 12 1 = 0 + = [0, a0, a1, a2, ..., an] 1 a0 + 1 a1 + 1 a2 + 1 a3 + 1 a4 + 1 ... + an q Conversely, if q < p, then is of the form p q 1 = 0 + (21) p 1 a0 + 1 a1 + 1 a2 + 1 a3 + 1 a4 + 1 ... + an and its reciprocal is p 1 1 = = a0 + q 1 1 a1 + 1 1 a0 + a2 + 1 1 a1 + a3 + 1 1 a2 + a4 + 1 1 a3 + ... + 1 an a4 + 1 ... + an (22) QED 13 This concludes the discussion of representing rational numbers by continued fractions. 14 3 Classical Theorems The following theorem is a nice way to calculate pk and qk by recursion formulae, just using the previous two terms. Theorem 3.1 Law of formation of the convergents For any k ≥ 2, pk = akpk−1 + pk−2 (23) qk = akqk−1 + qk−2 (24) Proof Proof by induction. When k = 2, it can be verified easily. Suppose they are valid for all k < n. Now consider the continued fraction [a1, a2, ..., an] and denote its 0 th pr r convergent by 0 .